CN104036456A - Image denoising apparatus and image denoising method - Google Patents
Image denoising apparatus and image denoising method Download PDFInfo
- Publication number
- CN104036456A CN104036456A CN201310073534.2A CN201310073534A CN104036456A CN 104036456 A CN104036456 A CN 104036456A CN 201310073534 A CN201310073534 A CN 201310073534A CN 104036456 A CN104036456 A CN 104036456A
- Authority
- CN
- China
- Prior art keywords
- denoising
- wave filter
- image
- input data
- orthogonal transformation
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Landscapes
- Image Processing (AREA)
Abstract
The invention provides an image denoising apparatus and an image denoising method. The image denoising apparatus comprises: an input unit, which is used for inputting data; a dimension number detection unit, which is used for detecting dimension numbers of the input data; a grouping unit, which is used for finding out a similar module in the input data; a filtering unit, which includes a multi-linear subs space learning filter and is used for carrying out denoising on the grouped input data; an aggregation unit, which is used for restoring the input data processed by denoising; and an output unit, which is used for outputting and displaying an image processed by denoising. The dimension number detection unit determines whether the input data are in a multi-dimensional mode; and if so, the multi-linear subs space learning filter carries out denoising processing. With the image denoising apparatus, denoising not only can be carried out on a two-dimensional image but also can be carried out a three-dimensional image or higher-dimensional images.
Description
Technical field
The present invention relates to a kind of image denoising device and image de-noising method for two dimension and multidimensional.
Background technology
Image denoising is very important preprocessing means.It can help improve image quality, improves image and cuts apart, the degree of accuracy of image registration etc.For medical image, can also assist doctor to carry out vision-based detection intuitively.Yet image denoising is also faced with huge challenge.Reason is that we will retain as far as possible all features of original image when removing noise.
In recent years, for two dimensional image, large quantities of denoising methods have been emerged, the BM3D method that wherein the most effectively the paper of Kostadin Dabov in 2006 " Image denoising with block-matching and 3D filtering " proposes relatively.For a width still image, the three-dimensional module forming for wherein similar two-dimensional block carries out space conversion, and carries out inverse transformation after utilizing shrinkage method to extract coefficient of efficiency, thereby reaches denoising object.
As shown in Figure 7, denoising step is as follows:
Step S701, system receives a new two-dimentional input picture;
Step S702, obtains three-dimensional module by the grouping based on two-dimensional block coupling;
Step S703, by orthogonal transformation wave filter denoising for this three-dimensional module;
Step S704, by the three-dimensional module polymerization after the denoising obtaining, obtains the two-dimentional output image after denoising.
Yet existing method is to take two dimensional image as denoising object, can not carry out denoising for three-dimensional and higher-dimension image.
Along with the development of modern technologies, the kind of image is also various all the more, such as two-dimentional natural image, and three-dimensional medical image, high-definition picture etc.And existing method can not judge the dimension of input picture, thereby can not use distinct methods to carry out denoising for different images.
Every width image has its different feature separately.Existing method, in carrying out space transfer process, uses one group of fixing conversion substrate to solve the Denoising Problems of all images.The variation that does not rely on input picture due to this group conversion substrate changes, and is difficult to the specific characteristic of corresponding different images.
In addition, in the process of space conversion, on every one dimension of existing method for the multidimensional module after forming, change respectively, do not consider the crosscorrelation contact between every one dimension.
Summary of the invention
Therefore the present invention completes in view of the above problems, and its object is to provide a kind of image denoising device, it is characterized in that possessing: input block, for inputting data; Dimension detecting unit, detects the dimension of described input data; Grouped element, for finding the similar module of described input data; Filter unit, comprises polyteny sub-space learning wave filter, for to through grouping described input data carry out denoising; Polymerized unit, for the described input data of reducing after denoising; And output unit, the image being used for after output display denoising, described dimension detecting unit judges whether described input data are multidimensional, if these input data are multidimensional, by described polyteny sub-space learning wave filter, carries out denoising.
According to such structure, can not only be to two dimensional image denoising, also can to three-dimensional even more higher-dimension image carry out denoising.And make full use of the information of input picture, the learning method of application self-adapting, considers the mutual relationship of image inside, improves denoising effect.
In addition, image denoising device of the present invention, it is characterized in that, also comprise demand estimation unit, for judging user's user demand, described filter unit also comprises the orthogonal transformation wave filter for input data are carried out to denoising, and described dimension detecting unit judgement input data are multidimensional or two dimension, if these input data are two dimensions, the rated condition based in described demand estimation unit is selected polyteny sub-space learning wave filter or orthogonal transformation wave filter.According to such structure, can judge the dimension of input picture, for different dimension images, adopt different denoising methods, can not only be to two dimensional image denoising, also can to three-dimensional even more higher-dimension image carry out denoising.
In addition, image denoising device of the present invention, is characterized in that, described rated condition is denoising quality exclusive disjunction speed.According to such structure, different demands that can corresponding practical application, for example, to time requirement and difference that denoising effect is required, realize denoising targetedly.
In addition, image denoising device of the present invention, is characterized in that, when described demand estimation unit is judged to be denoising quality when preferential, selects described polyteny sub-space learning wave filter; When described demand estimation unit is judged to be arithmetic speed when preferential, select described orthogonal transformation wave filter.According to such structure, different demands that can corresponding practical application, for example, to time requirement and difference that denoising effect is required, realize denoising targetedly.
In addition, image denoising device of the present invention, it is characterized in that, described input data are divided into two data, one of them data is processed by described polyteny sub-space learning wave filter, another data are processed by described orthogonal transformation wave filter, from the output data of described polyteny sub-space learning wave filter and merged from the output data of described orthogonal transformation wave filter.According to such structure, by conjunction with two kinds of different denoising methods, further improve denoising effect.
In addition, image denoising device of the present invention, it is characterized in that, described input data are carried out denoising by described orthogonal transformation wave filter in first step, in the situation of the result of being carried out denoising by described orthogonal transformation wave filter lower than predetermined value, in second step, by described polyteny sub-space learning wave filter, carry out denoising.According to such structure, by conjunction with two kinds of different denoising methods, further improve denoising effect.
In addition, image denoising device of the present invention, it is characterized in that, described polyteny sub-space learning wave filter is the wave filter based on GND-PCA, GND-ICA, MPCA or MICA, and described orthogonal transformation wave filter is discrete cosine transform, wavelet transform, discrete sine transform, discrete Fourier transformation, discrete Hartley transform or Walsh-Hadamard transform wave filter.According to such structure, can make full use of the information of input picture, the learning method of application self-adapting, considers the mutual relationship of image inside, improves denoising effect.
The image de-noising method that another object of the present invention is to provide a kind of image denoising device, is characterized in that, described image denoising device possesses: input block, for inputting data; Dimension detecting unit, detects the dimension of described input data; Grouped element, for finding the similar module of described input data; Filter unit, comprises polyteny sub-space learning wave filter, for to through grouping described input data carry out denoising; Polymerized unit, for the described input data of reducing after denoising; And output unit, the image being used for after output display denoising, described image de-noising method comprises: described dimension detecting unit judges whether described input data are multidimensional, if these input data are multidimensional, by described polyteny sub-space learning wave filter, carries out denoising.
According to such structure, can not only be to two dimensional image denoising, also can to three-dimensional even more higher-dimension image carry out denoising.And make full use of the information of input picture, the learning method of application self-adapting, considers the mutual relationship of image inside, improves denoising effect.
In addition, image de-noising method of the present invention, it is characterized in that, described image denoising device also comprises demand estimation unit, for judging user's user demand, described filter unit also comprises the orthogonal transformation wave filter for input data are carried out to denoising, described method comprises: described dimension detecting unit judgement input data are multidimensional or two dimension, if these input data are two dimensions, the rated condition based in described demand estimation unit is selected polyteny sub-space learning wave filter or orthogonal transformation wave filter.Thus, can judge the dimension of input picture, for different dimension images, adopt different denoising methods, can not only be to two dimensional image denoising, also can to three-dimensional even more higher-dimension image carry out denoising.
In addition, image de-noising method of the present invention, is characterized in that, described rated condition is denoising quality exclusive disjunction speed.Thus, different demands that can corresponding practical application, for example, to time requirement and difference that denoising effect is required, realize denoising targetedly.
In addition, image de-noising method of the present invention, is characterized in that, when described demand estimation unit is judged to be denoising quality when preferential, selects described polyteny sub-space learning wave filter; When described demand estimation unit is judged to be arithmetic speed when preferential, select described orthogonal transformation wave filter.Thus, different demands that can corresponding practical application, for example, to time requirement and difference that denoising effect is required, realize denoising targetedly.
In addition, image de-noising method of the present invention, it is characterized in that, described input data are divided into two data, one of them data is processed by described polyteny sub-space learning wave filter, another data are processed by described orthogonal transformation wave filter, from the output data of described polyteny sub-space learning wave filter and merged from the output data of described orthogonal transformation wave filter.Thus, by conjunction with two kinds of different denoising methods, further improve denoising effect.
In addition, image de-noising method of the present invention, it is characterized in that, described input data are carried out denoising by described orthogonal transformation wave filter in first step, in the situation of the result of being carried out denoising by described orthogonal transformation wave filter lower than predetermined value, in second step, by described polyteny sub-space learning wave filter, carry out denoising.Thus, by conjunction with two kinds of different denoising methods, further improve denoising effect.
In addition, image de-noising method of the present invention, it is characterized in that, described polyteny sub-space learning wave filter is the wave filter based on GND-PCA, GND-ICA, MPCA or MICA, and described orthogonal transformation wave filter is discrete cosine transform, wavelet transform, discrete sine transform, discrete Fourier transformation, discrete Hartley transform or Walsh-Hadamard transform wave filter.Thus, can make full use of the information of input picture, the learning method of application self-adapting, considers the mutual relationship of image inside, improves denoising effect.
As mentioned above, image denoising device of the present invention and image de-noising method have the following advantages:
(1) can judge the dimension of input picture, for different dimension images, adopt different denoising methods, can not only be to two dimensional image denoising, also can to three-dimensional even more higher-dimension image carry out denoising.
(2) according to the different demands of practical application, for example, to time requirement and the difference to denoising effect requirement, realize denoising targetedly;
(3) make full use of the information of input picture, the learning method of application self-adapting, considers the mutual relationship of image inside, improves denoising effect.
(4) by conjunction with two kinds of different denoising methods, further improve denoising effect.
Accompanying drawing explanation
Fig. 1 is the schematic diagram that uses whole MRI Medical Instruments of the present invention.
Fig. 2 is the structured flowchart of image denoising device of the present invention.
Fig. 3 is the process flow diagram of a denoising method of image denoising device of the present invention.
Fig. 4 is the process flow diagram of another denoising method of image denoising device of the present invention.
Fig. 5 is the process flow diagram of another denoising method of image denoising device of the present invention.
Fig. 6 is the process flow diagram of another denoising method of image denoising device of the present invention.
Fig. 7 is the process flow diagram of the image de-noising method of prior art.
Embodiment
Below in conjunction with drawings and Examples, the present invention will be described in more detail.
Referring to accompanying drawing, describe image denoising device involved in the present invention and image de-noising method in detail preferred embodiment.In addition, in the description of the drawings, to same or considerable part, be accompanied by prosign, the repetitive description thereof will be omitted.
The ultimate principle of image denoising device of the present invention and image de-noising method is described in conjunction with Fig. 1～Fig. 6.Wherein, Fig. 1 is the schematic diagram that uses whole MRI Medical Instruments of the present invention, and Fig. 2 is the structured flowchart of image denoising device of the present invention, and Fig. 3～Fig. 6 illustrates the method step of the denoising method of image denoising device of the present invention with process flow diagram.
Shown in Fig. 1, common MRI system consists of the following components: main magnet, Gradient Unit (gradient coil, gradient pulse program), radio frequency unit (radio-frequency coil, pulse protocol, receiving cable, transmission channel), computer system and spectrometer (computer memory, display), other utility appliance (shielding, sick bed).The present invention is directed to computer system and spectrometer improves, as shown in " image denoising " in Fig. 1 dotted line frame.Other parts are main magnet for example, Gradient Unit, and radio frequency unit and other utility appliance, within the present invention does not discuss.
As shown in Figure 2, image denoising device 1 of the present invention comprises input block 100, is used for receiving input noise image; Dimension detects list 200, is used for detecting the dimension of input noise image; Grouped element 300, for finding similar module; Demand estimation unit 400, is used for judging user's user demand; Filter unit 500, is used for carrying out noise-removed filtering; Polymerized unit 600, the image that is used for reducing after denoising; And output unit 700, be used for obtaining image after output display denoising.Wherein the form of the image of input is not limit, and can be DICOM, bmp, jpg, tiff, gif, the arbitrary formats such as raw.
As shown in Figure 3, for an input picture, its dimension uncertain, is likely two-dimentional image, is likely also three-dimensional or the image of higher-dimension more.Wherein, three-dimensional or the more dimension of the image of higher-dimension refer to the length of this image, wide, height, sample number and time etc.Dimension detects single 200 pairs of input data and first carries out dimension judgement.If two dimensional image, 300 pairs of grouped elements to reference to similar all of two-dimensional block, divide into groups, obtain a three-dimensional three-dimensional module.For this three-dimensional three-dimensional module, can adopt and carry out in two ways noise-removed filtering: a kind of is orthogonal transformation wave filter; Another kind is based on polyteny sub-space learning wave filter.If three-dimensional or high dimensional data, 300 pairs of grouped elements to reference to three-dimensional module or the similar all modules of higher-dimension module, divide into groups, obtain the module of a higher one dimension.Module for this higher one dimension mainly adopts based on polyteny sub-space learning wave filter.Finally, filtered module reverts to original position by polymerized unit 600 polymerizations, obtains denoising image.
Wherein, a kind of general conversion when orthogonal transformation wave filter refers to space conversion substrate, it does not change according to the variation of input picture, remain original transformation space, and this substrate can only be respectively converts respectively for every one dimension of the module after grouping, can not consider influencing each other between every one dimension; And refer to according to input picture based on polyteny sub-space learning wave filter, by considering influencing each other between every one dimension, at the bottom of system automatic learning goes out a transform-based, transformation space is changed because of the difference of input picture, thereby reach the object of auto adapted filtering.
Orthogonal transformation wave filter of the present invention, for example, can be discrete cosine transform (DCT), wavelet transform (DWT), discrete sine transform (DST), discrete Fourier transformation (DFT), discrete Hartley transform (DHT) or Walsh-Adama (Walsh-Hadamard) transformed filter etc.And of the present invention based on polyteny sub-space learning wave filter, be the wave filter based on GND-PCA (Generalized N Dimensional Principal Component Analysis), and be not limited to this kind of polyteny sub-space learning wave filter, other also can adopt as GND-ICA (Generalized N Dimensional Independent Component Analysis), MPCA or MICA etc.By using polyteny sub-space learning wave filter, thereby make the denoising of 3-D view become possibility.
Orthogonal transformation wave filter can not the various images of self-adaptation diversity, therefore, denoising effect is not as effective based on polyteny sub-space learning wave filter.But because orthogonal transformation wave filter is all the same to any image, need not relearn according to the difference of input picture, therefore, the processing time is relatively short based on polyteny sub-space learning wave filter.So, if denoising quality is preferential, select based on polyteny sub-space learning wave filter; If arithmetic speed is preferential, select orthogonal transformation wave filter.
Fig. 4 is the process flow diagram of another denoising method of explanation image denoising device 1 of the present invention.When demand estimation unit 400 is judged to be denoising quality when preferential, for two dimensional image, can apply based on polyteny sub-space learning wave filter and carry out denoising.By obtaining denoising image after polymerized unit 600 polymerizations.
Fig. 5 is the process flow diagram of another denoising method of explanation image denoising device 1 of the present invention.When demand estimation unit 400 is judged to be the preferential and denoising effect of denoising quality and has relatively high expectations, for two dimensional image, can apply orthogonal transformation wave filter and polyteny sub-space learning wave filter carries out denoising simultaneously.In polymerization process, filtered all of obtaining of the above-mentioned two kinds of methods of polymerized unit 600 application carry out polymerization simultaneously.
Fig. 6 is the process flow diagram of another denoising method of explanation image denoising device 1 of the present invention.When very high for denoising effect requirement.For two dimensional image, first with orthogonal transformation wave filter, image is carried out to denoising, after filtered polymerization, obtain image after the filtering of first stage.Afterwards, in the situation of the result of being carried out denoising by orthogonal transformation wave filter lower than predetermined value, using the filtered image of first stage as input picture, use again based on polyteny sub-space learning wave filter and carry out secondary filtering, 600 pairs filtered of polymerized unit after polymerization, obtains final image again.
As shown in Fig. 3～Fig. 6, for three-dimensional or high dimensional data, main employing carried out noise-removed filtering based on polyteny sub-space learning wave filter.
Display is exported the image after denoising after receiving above-mentioned information.
Below contrast Fig. 3～Fig. 6, set forth respectively the process that the present invention carries out denoising to two peacekeeping high dimensional datas.After receiving input data, all can carry out this process at every turn.
As shown in Figure 3:
Step S301, system receives a new input picture.
Step S302, system judges that whether the dimension of this input picture is higher than two dimension.
If be judged as "Yes" in step S302, follow-up processing is all carried out for higher-dimension image, performs step S303～S305, and is all the same method for Fig. 3～Fig. 6.
Step S303, in higher-dimension image
(dimension that wherein N is image, N>2, I
i, i=1,2 ..., N is the parts number on every one dimension), grouped element 300 for of this image with reference to higher-dimension module
find the higher-dimension module similar to it
after (m is the number of the similar higher-dimension piece that finds), grouping obtains the module of a higher one dimension
Step S304,500 pairs of filter units
module application is carried out denoising based on polyteny sub-space learning wave filter.Its process is as follows:
First, on every one dimension, by high dimensional data matrixing, calculate eigenvectors matrix.And according to eigenwert size and threshold size, the vectorial also composition characteristic vector matrix of former characteristics of correspondence that keeping characteristics value is relatively large.
Secondly, by alternative manner, find optimized eigenvectors matrix U on corresponding every one dimension
k, 1≤k≤N+1.Note, in the end one dimension, namely, in sample dimension, retains all proper vectors.
Again, by the module of higher one dimension
project on the space of the eigenvectors matrix generation after optimization:
Finally, by rebuilding, obtain the higher-dimension module after denoising:
Step S305, by the method for higher-dimension polymerization, polymerized unit 600 reverts to position original in image obtaining higher-dimension module after denoising, obtains denoising output image.The method of this higher-dimension polymerization is the expansion of the polymerization applied in prior art, makes it to high dimensional data, can carry out polymerization equally.And, in step S306, the data after denoising are exported.
If be judged as "No" in step S302, follow-up processing is all carried out for two dimensional image.And Fig. 3～Fig. 6 difference is said for two dimensional image.Therefore,, in following explanation, only in Fig. 3～Fig. 6, the dimension of system judgement input picture is not higher than two-dimentional situation.(for " dimension of input picture is higher than two dimension " be judged as "No" time)
As shown in Figure 3, when input picture is two dimensional image,
Step S313, grouped element 300 obtains three-dimensional module by grouping;
Step S314, it is preferential that demand estimation unit 400 is judged to be arithmetic speed;
Step S315, filter unit 500 is by orthogonal transformation wave filter denoising for this three-dimensional module;
Step S316, polymerized unit 600, by the three-dimensional module polymerization after the denoising obtaining, obtains the output image after denoising, and exports in step S306.
When input picture is two dimensional image, and demand estimation unit 400 is judged to be denoising effect when preferential, has following three kinds of situations, as Fig. 4, shown in Fig. 5 and Fig. 6.
Fig. 4 is the supplementary notes to Fig. 3.When requiring for denoising effect when high:
Step S413, in two dimensional image
(I
i, i=1,2 is the parts number on every one dimension), grouped element 300 for of this image with reference to two-dimensional block
find the two-dimensional block similar to it
after (m is the number of the similar higher-dimension piece that finds), grouping obtains the module of a higher one dimension
In step S414, it is preferential that demand estimation unit 400 is judged to be denoising effect;
Step S415,500 pairs of filter units
module application is carried out denoising based on polyteny sub-space learning wave filter.Its process is as follows:
First, on every one dimension, by high dimensional data matrixing, calculate eigenvectors matrix.And according to eigenwert size and threshold size, former proper vectors that keeping characteristics value is relatively maximum composition characteristic vector matrix.
Secondly, by alternative manner, find optimized eigenvectors matrix U on corresponding every one dimension
k, 1≤k≤3
Again, by three-dimensional module
project on the space of the eigenvectors matrix generation after optimization:
Finally, by rebuilding, obtain the three-dimensional module after denoising:
Step S416, by the method for higher-dimension polymerization, polymerized unit 600 reverts to position original in image obtaining higher-dimension module after denoising, obtains denoising output image.The method of this higher-dimension polymerization is the expansion of the polymerization applied in prior art, makes it to high dimensional data, can carry out polymerization equally.
Fig. 5 is the supplementary notes to Fig. 3.
Step S513, grouped element 300 obtains three-dimensional module by grouping;
Step S514, demand estimation unit 400 is judged to be denoising quality preferentially and denoising effect is had relatively high expectations;
Step S515, filter unit 500 is used respectively orthogonal transformation wave filter and the wave filter denoising of polyteny subspace by this three-dimensional module simultaneously;
Step S516, obtains three-dimensional module polymerization together after the denoising that polymerized unit 600 obtains two kinds of wave filters simultaneously, obtain denoising output image.
Fig. 6 is the supplementary notes to Fig. 3.When demand estimation unit 400 is judged to be denoising effect requirement when the highest:
Step S613, grouped element 300 obtains three-dimensional module by grouping;
Step S615, filter unit 500 is by orthogonal transformation wave filter denoising for this three-dimensional module;
Step S616, polymerized unit 600, by the three-dimensional module polymerization after the denoising obtaining, obtains the output image after denoising for the first time.But, because the result of carrying out denoising by orthogonal transformation wave filter is lower than predetermined value, therefore also to carry out denoising for the second time.
At step S617, grouped element 300 obtains three-dimensional module by grouping again by the image after denoising for the first time
At step S618,500 pairs of three-dimensional module application polyteny sub-space learning wave filters of filter unit carry out denoising (step is described consistent with Fig. 3).
At step S619, by the method by polymerized unit 600 polymerizations, obtaining three-dimensional module after denoising, revert to position original in image, obtain final denoising image, and export in step S620.
According to above-mentioned each denoising method, can judge the dimension of input picture, for different dimension images, adopt different denoising methods, can not only be to two dimensional image denoising, also can to three-dimensional even more higher-dimension image carry out denoising, and according to the different demands of practical application, for example, to time requirement and the difference to denoising effect requirement, realize denoising targetedly.
Although below in conjunction with the accompanying drawings and embodiments the present invention is illustrated, is appreciated that above-mentioned explanation does not limit the present invention in any form.For example, in image denoising device of the present invention, input data can be divided into two data, one of them data is processed by polyteny sub-space learning wave filter, another data are processed by orthogonal transformation wave filter, then by the output data from described polyteny sub-space learning wave filter with from the output data of described orthogonal transformation wave filter, merge.Thus, by conjunction with two kinds of different denoising methods, further improve denoising effect.
Those skilled in the art can be out of shape and change the present invention as required in the situation that not departing from connotation of the present invention and scope, and these distortion and variation all fall within the scope of the present invention.
Claims (14)
1. an image denoising device, is characterized in that,
Possess:
Input block, for inputting data;
Dimension detecting unit, detects the dimension of described input data;
Grouped element, for finding the similar module of described input data;
Filter unit, comprises polyteny sub-space learning wave filter, for to through grouping described input data carry out denoising;
Polymerized unit, for the described input data of reducing after denoising; And
Output unit, the image being used for after output display denoising,
Described dimension detecting unit judges whether described input data are multidimensional, if these input data are multidimensional, by described polyteny sub-space learning wave filter, carries out denoising.
2. image denoising device as claimed in claim 1, is characterized in that,
Also comprise demand estimation unit, for judging user's user demand,
Described filter unit also comprises the orthogonal transformation wave filter for input data are carried out to denoising,
Described dimension detecting unit judgement input data are multidimensional or two dimension, if these input data are two dimensions, the rated condition based in described demand estimation unit is selected polyteny sub-space learning wave filter or orthogonal transformation wave filter.
3. image denoising device as claimed in claim 2, is characterized in that,
Described rated condition is denoising quality exclusive disjunction speed.
4. image denoising device as claimed in claim 3, is characterized in that,
When described demand estimation unit is judged to be denoising quality when preferential, select described polyteny sub-space learning wave filter;
When described demand estimation unit is judged to be arithmetic speed when preferential, select described orthogonal transformation wave filter.
5. image denoising device as claimed in claim 2, is characterized in that,
Described input data are divided into two data, and one of them data is processed by described polyteny sub-space learning wave filter, and another data are processed by described orthogonal transformation wave filter,
From the output data of described polyteny sub-space learning wave filter and merged from the output data of described orthogonal transformation wave filter.
6. image denoising device as claimed in claim 2, is characterized in that,
Described input data are carried out denoising by described orthogonal transformation wave filter in first step, in the situation of the result of being carried out denoising by described orthogonal transformation wave filter lower than predetermined value, in second step, by described polyteny sub-space learning wave filter, carry out denoising.
7. the image denoising device as described in any one in claim 1～6, is characterized in that,
Described polyteny sub-space learning wave filter is the wave filter based on GND-PCA, GND-ICA, MPCA or MICA, and described orthogonal transformation wave filter is discrete cosine transform, wavelet transform, discrete sine transform, discrete Fourier transformation, discrete Hartley transform or Walsh-Hadamard transform wave filter.
8. an image de-noising method for image denoising device, is characterized in that,
Described image denoising device possesses:
Input block, for inputting data;
Dimension detecting unit, detects the dimension of described input data;
Grouped element, for finding the similar module of described input data;
Filter unit, comprises polyteny sub-space learning wave filter, for to through grouping described input data carry out denoising;
Polymerized unit, for the described input data of reducing after denoising; And
Output unit, the image being used for after output display denoising,
Described image de-noising method comprises:
Described dimension detecting unit judges whether described input data are multidimensional, if these input data are multidimensional, by described polyteny sub-space learning wave filter, carries out denoising.
9. image de-noising method as claimed in claim 8, is characterized in that,
Described image denoising device also comprises demand estimation unit, for judging user's user demand,
Described filter unit also comprises the orthogonal transformation wave filter for input data are carried out to denoising,
Described method comprises: described dimension detecting unit judgement input data are multidimensional or two dimension, if these input data are two dimensions, the rated condition based in described demand estimation unit is selected polyteny sub-space learning wave filter or orthogonal transformation wave filter.
10. image de-noising method as claimed in claim 9, is characterized in that,
Described rated condition is denoising quality exclusive disjunction speed.
11. image de-noising methods as claimed in claim 10, is characterized in that,
When described demand estimation unit is judged to be denoising quality when preferential, select described polyteny sub-space learning wave filter;
When described demand estimation unit is judged to be arithmetic speed when preferential, select described orthogonal transformation wave filter.
12. image de-noising methods as claimed in claim 9, is characterized in that,
Described input data are divided into two data, and one of them data is processed by described polyteny sub-space learning wave filter, and another data are processed by described orthogonal transformation wave filter,
From the output data of described polyteny sub-space learning wave filter and merged from the output data of described orthogonal transformation wave filter.
13. image de-noising methods as claimed in claim 9, is characterized in that,
Described input data are carried out denoising by described orthogonal transformation wave filter in first step, in the situation of the result of being carried out denoising by described orthogonal transformation wave filter lower than predetermined value, in second step, by described polyteny sub-space learning wave filter, carry out denoising.
14. image de-noising methods as described in any one in claim 8～13, is characterized in that,
Described polyteny sub-space learning wave filter is the wave filter based on GND-PCA, GND-ICA, MPCA or MICA, and described orthogonal transformation wave filter is discrete cosine transform, wavelet transform, discrete sine transform, discrete Fourier transformation, discrete Hartley transform or Walsh-Hadamard transform wave filter.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310073534.2A CN104036456A (en) | 2013-03-08 | 2013-03-08 | Image denoising apparatus and image denoising method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310073534.2A CN104036456A (en) | 2013-03-08 | 2013-03-08 | Image denoising apparatus and image denoising method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN104036456A true CN104036456A (en) | 2014-09-10 |
Family
ID=51467217
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201310073534.2A Pending CN104036456A (en) | 2013-03-08 | 2013-03-08 | Image denoising apparatus and image denoising method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104036456A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106875280A (en) * | 2017-03-02 | 2017-06-20 | 深圳大图科创技术开发有限公司 | Integrated community service platform |
| CN113820016A (en) * | 2021-08-19 | 2021-12-21 | 东南大学 | Phase modulation thermal wave signal total variation denoising method |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6018561A (en) * | 1998-07-27 | 2000-01-25 | Siemens Corporate Research, Inc. | Mask boundary correction in a cone beam imaging system using simplified filtered backprojection image reconstruction |
| US20070110294A1 (en) * | 2005-11-17 | 2007-05-17 | Michiel Schaap | Image enhancement using anisotropic noise filtering |
| CN101582984A (en) * | 2009-04-14 | 2009-11-18 | 公安部物证鉴定中心 | Method and device for eliminating image noise |
| CN102203826A (en) * | 2008-12-25 | 2011-09-28 | 梅迪奇视觉成像解决方案有限公司 | Denoising medical images |
| CN102682429A (en) * | 2012-04-13 | 2012-09-19 | 泰山学院 | De-noising method of filtering images in size adaptive block matching transform domains |
-
2013
- 2013-03-08 CN CN201310073534.2A patent/CN104036456A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6018561A (en) * | 1998-07-27 | 2000-01-25 | Siemens Corporate Research, Inc. | Mask boundary correction in a cone beam imaging system using simplified filtered backprojection image reconstruction |
| US20070110294A1 (en) * | 2005-11-17 | 2007-05-17 | Michiel Schaap | Image enhancement using anisotropic noise filtering |
| CN102203826A (en) * | 2008-12-25 | 2011-09-28 | 梅迪奇视觉成像解决方案有限公司 | Denoising medical images |
| CN101582984A (en) * | 2009-04-14 | 2009-11-18 | 公安部物证鉴定中心 | Method and device for eliminating image noise |
| CN102682429A (en) * | 2012-04-13 | 2012-09-19 | 泰山学院 | De-noising method of filtering images in size adaptive block matching transform domains |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106875280A (en) * | 2017-03-02 | 2017-06-20 | 深圳大图科创技术开发有限公司 | Integrated community service platform |
| CN113820016A (en) * | 2021-08-19 | 2021-12-21 | 东南大学 | Phase modulation thermal wave signal total variation denoising method |
| CN113820016B (en) * | 2021-08-19 | 2022-11-11 | 东南大学 | Phase modulation thermal wave signal total variation denoising method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US9858472B2 (en) | Three-dimensional facial recognition method and system | |
| CN102395037B (en) | Format recognition method and device | |
| CN101221661B (en) | Image registration method and device | |
| Routray et al. | Efficient hybrid image denoising scheme based on SVM classification | |
| Tiwari et al. | Designing palmprint based recognition system using local structure tensor and force field transformation for human identification | |
| CN104240187A (en) | Image denoising device and image denoising method | |
| CN105809182B (en) | Image classification method and device | |
| Kim et al. | Exact order based feature descriptor for illumination robust image matching | |
| Zhang et al. | Steganalysis of content-adaptive JPEG steganography based on Gauss partial derivative filter bank | |
| CN104036456A (en) | Image denoising apparatus and image denoising method | |
| CN113837310A (en) | Multi-scale fusion certificate copying and identifying method and device, electronic equipment and medium | |
| WO2017112087A1 (en) | Reconstruction of signals using a gramian matrix | |
| CN102013101A (en) | Blind detection method of permuted and tampered images subjected to fuzzy postprocessing | |
| He et al. | Decomposition for a quaternion tensor triplet with applications | |
| KR20210030628A (en) | Contents recommendation system using deep-learning and style transfer | |
| CN104282012A (en) | Wavelet domain based semi-reference image quality evaluating algorithm | |
| Bi et al. | A robust color edge detection algorithm based on the quaternion Hardy filter | |
| Zhang et al. | A highly repeatable feature detector: improved Harris–Laplace | |
| ITVI20120041A1 (en) | DETECTION OF CHARACTERISTICS OF AN IMAGE | |
| CN114239858B (en) | Graph learning method and device for distributed graph model | |
| CN102881008A (en) | Circular loop statistic characteristic-based anti-rotation image Hash method | |
| Sur et al. | An a contrario model for matching interest points under geometric and photometric constraints | |
| TW201909116A (en) | Depth calculation method and device thereof | |
| Xie et al. | Efficient rotation-and scale-invariant texture classification method based on Gabor wavelets | |
| CN106778772B (en) | A kind of significant extracting method of stereo-picture vision |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20140910 |
|
| WD01 | Invention patent application deemed withdrawn after publication |